Markov State Models package (pyemma.msm)

The msm package provides functions to estimate, analyze and generate discrete-state Markov models. All public functions accept dense NumPy and sparse SciPy matrices and automatically choose the corresponding implementation.

User Functions

For most users, the following high-level functions are sufficient to estimate msm models from data. Expert users may want to construct Estimators or Models (see below) directly.

markov_model(P[, dt_model]) Markov model with a given transition matrix
timescales_msm(dtrajs[, lags, nits, ...]) Implied timescales from Markov state models estimated at a series of lag times.
its(dtrajs[, lags, nits, reversible, ...]) Implied timescales from Markov state models estimated at a series of lag times.
estimate_markov_model(dtrajs, lag[, ...]) Estimates a Markov model from discrete trajectories
bayesian_markov_model(dtrajs, lag[, ...]) Bayesian Markov model estimate using Gibbs sampling of the posterior
tpt(msmobj, A, B) A->B reactive flux from transition path theory (TPT)
timescales_hmsm(dtrajs, nstates[, lags, ...]) Calculate implied timescales from Hidden Markov state models estimated at a series of lag times.
estimate_hidden_markov_model(dtrajs, ...[, ...]) Estimates a Hidden Markov state model from discrete trajectories
bayesian_hidden_markov_model(dtrajs, ...[, ...]) Bayesian Hidden Markov model estimate using Gibbs sampling of the posterior

MSM classes

Estimators to generate models from data. If you are not an expert user, use the API functions above.

ImpliedTimescales(estimator[, lags, nits, ...]) Implied timescales for a series of lag times.
ChapmanKolmogorovValidator(model, estimator, ...)
MaximumLikelihoodMSM([lag, reversible, ...]) Maximum likelihood estimator for MSMs given discrete trajectory statistics
BayesianMSM([lag, nsamples, nsteps, ...]) Bayesian Markov state model estimator
MaximumLikelihoodHMSM([nstates, lag, ...]) Maximum likelihood estimator for a Hidden MSM given a MSM
BayesianHMSM([nstates, lag, stride, ...]) Estimator for a Bayesian Hidden Markov state model

Models of the kinetics or stationary properties of the data. If you are not an expert user, use the API functions above.

MSM(P[, pi, reversible, dt_model, neig, ncv]) Markov model with a given transition matrix
SampledMSM(samples[, ref, conf]) Sampled Markov state model
HMSM(P, pobs[, pi, dt_model]) Hidden Markov model on discrete states.
SampledHMSM(samples[, ref, conf]) Sampled Hidden Markov state model
ReactiveFlux(A, B, flux[, mu, qminus, ...]) A->B reactive flux from transition path theory (TPT)
PCCA(P, m) PCCA+ spectral clustering method with optimized memberships [1]_